Professional Documents
Culture Documents
179
JPE 11-2-9
Accurate MATLAB Simulink PV System Simulator
Based on a Two-Diode Model
Kashif Ishaque
, Zainal Salam
Dept. of Energy Conversion, Faculty of Electrical Engineering, Universiti Teknologi Malaysia, Johor, Malaysia
Abstract
This paper proposes a MATLAB Simulink simulator for photovoltaic (PV) systems. The main contribution of this work is
the utilization of a two-diode model to represent a PV cell. This model is known to have better accuracy at low irradiance levels
which allows for a more accurate prediction of PV system performance. To reduce computational time, the input parameters are
reduced to four and the values of R
p
and R
s
are estimated by an efcient iteration method. Furthermore, all of the inputs to the
simulator are information available on a standard PV module datasheet. The simulator supports large array simulations that can
be interfaced with MPPT algorithms and power electronic converters. The accuracy of the simulator is veried by applying the
model to ve PV modules of different types (multi-crystalline, mono-crystalline, and thin-lm) from various manufacturers. It is
envisaged that the proposed work can be very useful for PV professionals who require a simple, fast and accurate PV simulator
to design their systems.
Key Words: Mono-crystalline, Multi-crystalline, PV module, STC, Thin-lm
I. INTRODUCTION
Large and small scale PV power systems have been com-
mercialized in many countries due to their potential long term
benets, generous fed-in tariff schemes and other attractive
initiatives provided by various governments to promote sus-
tainable green energy. In PV power generation, due to the
high cost of modules, optimal utilization of the available
solar energy has to be ensured. This mandates an accurate
and reliable simulation of designed PV systems prior to
installation.
The most important component that affects the accuracy of a
simulation is the PV cell modeling, which primarily involves
the estimation of the non-linear I-V and P-V characteristics
curves. Though impractical, the simplest model is the single
diode model i.e. a current source in parallel to a diode [1]-
[4]. It only requires three parameters, namely the short-circuit
current (I
sc
), the open circuit voltage (V
oc
) and the diode
ideality factor (a). This model is improved by the inclusion of
one series resistance, R
s
[5]-[10], as shown in Fig.1(a). Despite
its simplicity, this model exhibits serious deciencies when
subjected to temperature variations. An extension of the model
which includes an additional shunt resistance R
p
is shown
in Fig.1 (b) [11]-[15]. Although a signicant improvement is
achieved, this model demands signicant computational effort.
Furthermore its accuracy deteriorates at low irradiance levels,
Manuscript received Jun. 5, 2010; revised Dec. 9, 2010
1
T
)] (3)
where E
g
is the band gap energy of the semiconductor and
I
0,STC
is the nominal saturation current. An improved equation
to describe the saturation current which considers temperature
variations is given by [15]:
I
0
=
(I
sc STC
+K
I
T)
exp[(V
oc,STC
+K
V
T)/aV
T
] 1
. (4)
Accurate MATLAB Simulink PV System Simulator Based on . . . 181
The constant K
v
is the open circuit voltage coefcient. This
value is available from datasheets. To further simplify the
model, in this work, both of the reverse saturation currents,
I
o1
and I
o2
are set be to equal in magnitude.
I
01
=I
02
=
(I
sc STC
+K
I
T)
exp[(V
oc,STC
+K
V
T)/{(a
1
+a
2
)/p}V
T
] 1
. (5)
The equalization simplies the computation as no itera-
tion is required. As a result, the solution can be obtained
analytically. The diode ideal factors a
1
and a
2
represent the
diffusion and recombination current components, respectively.
In accordance with Shockleys diffusion theory, the diffusion
current, a
1
must be unity [16]. The value of a
2
, however, is
exible. Based on the extensive simulations that were carried
out, it was found that if a
2
1.2, the best match between the
proposed model and the practical I-V curve is obtained. Since
(a
1
+a
2
)/p = 1 and a
1
= 1, it follows that the variable p can
be chosen to be 2.2. The following expression for I
o1
and
I
o2
results in:
I
01
= I
02
=
(I
sc STC
+K
I
T)
exp[(V
oc,n
+K
V
T)/V
T
] 1
. (6)
This generalization can eliminate the ambiguity in selecting
the values of a
1
and a
2
. Using (2) and (5), the ve parameters
of this model can be readily determined, i.e. I
PV
, I
o1
, I
o2
, a
1
and a
2
. Furthermore, the iteration process to compute I
01
and
I
02
is avoided, resulting in a reduced computing time.
2) Determination of the R
p
and R
s
Values: The remaining
two parameters, i.e. R
p
and R
s
are obtained through iteration.
Several researchers have estimated these two parameters inde-
pendently, but the results are unsatisfactory. In this work, R
p
and R
s
are calculated simultaneously, similar to the procedure
proposed in [15]. This approach has not been applied to a
two-diode model. The idea is to maximize the power point
(P
mp
) matching; i.e. to match the calculated peak power (P
mp,C
)
and the experimental (from the manufacturers datasheet) peak
power (P
mp,E
) by iteratively increasing the value of R
s
while
simultaneously calculating the value of R
p
. From (1) under the
maximum power point condition, the expression of R
p
can be
rearranged and rewritten as:
R
p
=
V
mp
(V
mp
+I
m
R
s
)
[V
mp
{I
PV
I
d1
I
d2
}P
max,E
]
(7)
where
I
d1
= I
01
[exp(
V +IRs
a
1
V
T1
) 1] (8)
and
I
d2
= I
02
[exp(
V +IR
s
a
2
V
T2
) 1]. (9)
The initial conditions for both resistances are given below:
R
so
= 0; R
po
=
V
mp
I
scn
I
mp
V
ocn
V
mp
I
mp
(10)
Fig. 3 depicts the mechanism of the iteration to obtain
the correct R
s
value. Two types of PV modules, Kyocera
KC200GT [28] and Solarex MSX-60 [29] were chosen for
illustration. In every case, R
s
is increased until P
max,C
becomes
exactly equal to P
max,E
. Meanwhile, for each iteration, the
value of R
p
is calculated simultaneously using (7).
Fig. 3. Matching P-V curves methodology for three PV modules.
Fig. 4. Matching I-V curves for KC200GT and MSX-60 modules.
The nal matched I-V curves for the two PV modules are
shown in Fig. 4. As can be observed the three important points,
namely I
sc,
, P
mp
(V
mp
, I
mp
) and V
oc
strongly agree with the
manufacturer data. With the availability of all seven param-
eters, the output current of the cell can now be determined
using the standard Newton-Raphson method. The owchart
that describes the P
mp
matching algorithm is given in Fig. 5.
C. Large Array Simulation
In a typical installation of a large PV power generation
system, the modules are congured in a series parallel struc-
ture (i.e. N
ss
N
pp
modules), as depicted in Fig. 6. To handle
such cases, the output current equation given in (11) has to be
modied as follows:
I = I
PV
N
pp
I
01
N
pp
_
_
exp
_
_
_
_
V +IR
s
_
N
ss
N
pp
_
a
1
V
T
N
ss
_
_
_
_
1
_
I
02
N
pp
_
_
exp
_
_
_
_
V +IR
s
_
N
ss
N
pp
_
a
2
V
T
N
ss
_
_
_
_
1
_
_
_
_
_
V +IR
s
_
N
ss
N
pp
_
R
p
_
N
ss
N
pp
_
_
_
_
_
(11)
where I
PV
, I
01
, I
02
, R
p
, R
s
, a
1
, a
2
are the parameters of the indi-
vidual modules.
Fig. 7 shows a building block of a PV array for any number
of series parallel connections. The inputs for a two diode
model are designed as a template shown in Fig. 8 (a). All of the
parameters for the inputs are available from the manufactures
182 Journal of Power Electronics, Vol. 11, No. 2, March 2011
Fig. 5. Matching Algorithm.
Fig. 6. Series parallel combination in PV array.
datasheet. A mask implementation two diode PV simulator is
shown in Fig. 8 (b).
III. RESULTS AND DISCUSSION
A. Verication of the Two Diode Model
The two diode model described in this paper is validated
by the measured parameters of selected PV modules. Five
modules of different brands/models are utilized for verica-
tion; these include the multi- and mono-crystalline as well
as the thin-lm types. The specications of the modules
are summarized in Table I. The computational results are
compared with the R
s
[6] and R
p
[15] models. Note that these
two models are shown in Figs. 1(a) and (b), respectively.
Table II shows the parameters of the proposed two-diode
model. Although it has more variables, the actual number of
parameters to be computed is only four because I
o1
=I
o1
while
a
1
and a
2
can be chosen arbitrarily from (5).
Fig. 9 shows the I-V curves for a KC200GT module (single),
for different levels of irradiation (per unit quantity: Sun=1
equivalent to 1000W/m2). The calculated values from the
proposed two-diode and R
p
- models are evaluated against the
measured data from the manufacturers datasheet. Comparison
with the R
s
-model is not included to avoid overcrowding of the
plot. However, the results for the R
s
-model will be analyzed
later in the performance evaluation of the three models.
Fig. 7. PV simulator Block in Simulink.
(a) (b)
Fig. 8. (a) Input parameter window. (b) mask implementation of PV simulator.
The proposed two-diode model and the R
p
-model exhibit
similar results under STC. This is to be expected because
both models use similar maximum power matching algorithms
to evaluate the model parameters under STC. However, as
irradiance gets lower, more accurate results are obtained
from the two-diode model, especially in the vicinity of the
open circuit voltage. At V
oc
, the R
p
-model shows a departure
from the experimental data, suggesting that the R
p
-model is
inadequate when dealing with low irradiance levels. This is
envisaged to have signicant implications during periods of
partial shading.
The performance of the models when subjected to temper-
ature variations is considered next. All of measurements are
conducted under the STC irradiance of 1000W/m2. The pro-
posed model is compared with the R
s
-model. This comparison
was specically chosen to highlight the signicant problems
with the R
s
-model when subjected to temperature variations.
The R
p
-model is not shown for the sake of simplicity, but
it will be included later in an analysis that compares all
three models together. Two modules are tested, namely the
KC200GT and the MSX-60. As can be seen in Figs. 10
and 11, respectively, the I-V curves computed by the two-
diode model accurately match the experimental data for all
Accurate MATLAB Simulink PV System Simulator Based on . . . 183
TABLE I
TABLE ISTC SPECIFICATIONS FOR THE THREE MODULES USED IN THE EXPERIMENTS
Parameter Multi-Crystalline Mono-Crystalline Thin-Film
BP Solar
MSX-60
Kyocera
KG200GT
Shell
S36
Shell
SP-70
Shell
ST40
I
sc
3.8 A 8.21 A 2.3 A 4.7 A 2.68 A
V
oc
21.1 V 32.9 V 21.4 V 21.4 V 23.3 V
I
mp
3.5 A 7.61 A 2.18 A 4.25 A 2.41 A
V
mp
17.1V 26.3 V 16.5 V 16.5 V 16.6 V
K
v
80 mV/
C 123 mV/
C 76 mV/
C 76 mV/
C 100 mV/
C
K
i
3 mA/oC 3.18 mA/
C 1 mA/
C 2 mA/
C 0.35 mA/
C
N
s
36 54 36 36 36
TABLE II
PARAMETERS FOR THE PROPOSED TWO-DIODE MODEL
Parameter Multi-Crystalline Mono-Crystalline Thin-Film
BP Solar
MSX-60
Kyocera
KG200GT
Shell
S36
Shell
SP-70
Shell
ST40
I
sc
3.8 A 8.21 A 2.3 A 4.7 A 2.68 A
V
oc
21.1 V 32.9 V 21.4 V 21.4 V 23.3 V
I
mp
3.5 A 7.61 A 2.16 A 4.24 A 2.41 A
V
mp
17.1V 26.3 V 16.7 V 16.5 V 16.6 V
I
o1
= I
o2
4.70410
10
A 4.21810
10
A 2.05910
10
A 4.20610
10
A 3.0710
11
A
I
PV
3.80 A 8.21 A 2.3 A 4.7 A 2.68A
R
p
R
s
176.4
0.35
160.5
0.32
806.4
0.89
91
0.51
211.7
1.76
Fig. 9. I-V curves of R
p
-Model and proposed two-diode model of the
KC200GT PV module for several Irradiation levels.
Fig. 10. I-V curves of R
s
and proposed two-diode model of the KC200GT
PV module for several temperature levels.
temperature conditions. In contrast, at higher temperature, the
results from the R
s
-model deviate from the measured values
quite signicantly.
Fig. 12 shows an analyses of the relative error of V
oc
and
the maximum power point P
mp
for a KC200GT module at
Fig. 11. I-V curves of R
s
and proposed two-diode model of the MSX 60 PV
module for several temperature levels @ 1KW/m
2
.
different irradiance levels. The temperature is set to STC.
The relative error is dened as the difference between the
simulated and the measured V
oc
and P
mp
values. The difference
is then divided by the measured value. As can be seen, under
the STC irradiance level, there is a very small difference
in the V
oc
values among the three models. However as the
irradiance is reduced, signicant deviations are observed with
the R
s
and R
p
-models. Similar results can be seen for the P
mp
.
On the other hand, the proposed two-diode model accurately
calculates P
mp
under all irradiance levels.
Fig. 13 shows the performance of the three models when
subjected to variations in module temperature. The irradiance
is set to STC. There is no signicant difference between the
R
p
and the two-diode models. However the R
s
model exhibits
poor performance in both V
oc
and P
mp
calculations.
To show the effectiveness of the models for different silicon
technologies, comparisons between a S36 (mono-crystalline),
a SP70 (multi-crystalline), and a ST40 (thin-lm) are carried
out. All of these modules where manufactured by Shell
[30]. For this test, the irradiance is maintained constant at
184 Journal of Power Electronics, Vol. 11, No. 2, March 2011
(a)
(b)
Fig. 12. Relative error for V
oc
and P
mp
, for R
s
, R
p
and the proposed two-diode
model for KC200GT PV module.
(a)
(b)
Fig. 13. Relative error for V
oc
and P
mp
, for R
s
, R
p
and the proposed two-diode
model for KC200GT PV module.
STC. Fig. 14 shows the relative error in P
mp
for a wide
variation in temperature (25
C to +50